An Inspection-Based Compositional Approach to the Quantitative Evaluation of Assembly Lines

We present a model-based approach to performance evaluation of a collection of similar systems based on runtime observations. As a concrete example, we consider an assembly line made of sequential workstations with transfer blocking and no buffering capacity, implementing complex workflows with random choices and sequential/cyclic phases with generally distributed durations and no internal parallelism. Starting from the steady state, an inspection mechanism is subject to some degree of uncertainty in the identification of the current phase of each workstation, and is in any case unable to estimate remaining times. By relying on the positive correlation between delays at different workstations, we provide stochastic upper and lower approximations of the performance measures of interest, including the time to completion of the local workflow of each workstation and the time until when a workstation starts a new job. Experimental results show that the approximated evaluation is accurate and feasible for lines of significant complexity.